Administering a Cluster

Creating the kubectl config

kubectl is the command line client used to control a kubernetes cluster. Kubernetes uses a separate authentication mechanism from OpenStack (Keystone). Since OpenStack already knows who we are, we can use it to generate a kubectl config file that includes credentials to authenticate us to Kubernetes.

1. Create a new directory and cd to it;

   mkdir mycluster
   cd mycluster
   

2. Use the openstack tool to generate the config file for kubectl:

   openstack coe cluster config mycluster
   export KUBECONFIG=/home/jake/temp/coe/mycluster/config
   

3. Set the KUBECONFIG environment variable used by copy-and-pasting the export KUBECONFIG=... line output by the above to the shell prompt:

   export KUBECONFIG=/home/jake/temp/coe/mycluster/config
   echo $KUBECONFIG
   /home/jake/temp/coe/mycluster/config
   

Using kubectl

1. Use kubectl to see if all your service pods in kubernetes are set up correctly. All the pods in the kube-system namespace should have status Running. For example:

   kubectl get all --all-namespaces
   

   NAMESPACE                NAME                                                                 READY   STATUS      RESTARTS       AGE
   calico-apiserver         pod/calico-apiserver-79dbf75bfc-997sh                                1/1     Running     0              45d
   calico-apiserver         pod/calico-apiserver-79dbf75bfc-wjqsz                                1/1     Running     0              45d
   calico-system            pod/calico-kube-controllers-5c7ffcfbf9-q75rz                         1/1     Running     0              45d
   calico-system            pod/calico-node-49dsv                                                1/1     Running     2 (45d ago)    45d
   calico-system            pod/calico-node-q69tj                                                1/1     Running     0              45d
   calico-system            pod/calico-typha-6c874f96b5-7w7pv                                    1/1     Running     0              45d
   calico-system            pod/csi-node-driver-hl9jm                                            2/2     Running     0              45d
   calico-system            pod/csi-node-driver-jqwrv                                            2/2     Running     0              45d
   kube-system              pod/coredns-7c65d6cfc9-fzqmp                                         1/1     Running     0              45d
   kube-system              pod/coredns-7c65d6cfc9-n7l9p                                         1/1     Running     0              45d
   [output truncated]
   

Using the web interface

Magnum used to ship its own Web UI Dashboard, with Magnum installing it by default. This is now deprecated and no-longer maintained, headlamp is now the recommended application for this.

Headlamp gives you a browser-based view of your cluster’s pods, deployments, services, logs and more.

Installation instructions can be found on the Headlamp site.

Upgrading a cluster

You can upgrade your cluster to a newer Kubernetes version from the Nectar Dashboard. The upgrade is a rolling upgrade, meaning nodes are replaced one at a time so your cluster keeps running while it upgrades.

1. Go to Project > Container Infra > Clusters. Find your cluster in the list, open the dropdown menu next to Manage Node Groups and select Rolling Cluster Upgrade.

   

2. In the Rolling Cluster Upgrade dialog, select the Kubernetes version to upgrade to and click Submit. For example, a cluster running v1.34.7 can be upgraded to 1.35.4.

   

The cluster status changes to UPDATE_IN_PROGRESS while nodes are replaced, and returns to UPDATE_COMPLETE when the upgrade has finished. You can watch the nodes being replaced with kubectl get nodes.

Upgrade one version at a time
Kubernetes control plane upgrades on Nectar must proceed through each nominated release in order; skipping versions is not supported.